Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
USB 3.x PCIe Cards for Classic Mac Pro
- Thread starter ActionableMango
- WikiPost WikiPost
- Start date
- Sort by reaction score
- Status
The new Sonnet cards' "resettable fuses" seems like a good feature other/cheaper cards won't have. Anyone have any (technical) thoughts on this?
It's commonly used. From RaspberryPi to Mac Pro USB ports lots of designs use polyfuses to protect the circuit. Nothing new here.
They state: Unlike ordinary USB-C cards, the Allegro card features a resettable fuse on each port, protecting both card and computer should a cable or peripheral fail.
So you're saying Sonnet is selling snake oil here?
I do hesitate at what kind of protection a $25 card will have for my 5,1 motherboard...
I can't say anything without knowing the circuit, but it's at least 10 year old tech, the first RaspeberryPi had it and MP5,1 has lots of poly fuses too. Maybe they have something better? Just using one for each port is nothing new.
Anyway, Mac Pro has one for each port. Nowadays, it's a cheap way to protect the USB ports.
https://www.digikey.com/Web Export/..._8004/PDF/te-appnote-usb2-usb3-protection.pdf
Last edited:
I wonder if they know something we don't about those "ordinary" USB-C cards? Perhaps those older USB 3.0 poly fuse specs don't live up to the higher current draw of USB-C?
The article from Digikey that I linked has PPTC adequate for the USB-C normal amperage, charging ports like the 85W MacBook Pro ports need a beefier solution.
"Sonnet’s Allegro USB-C is unique in regulating power to each port independently, giving each one protection from power glitches on the adjacent port. Without this protection (such as found on an ordinary card), connects and disconnects of one device may cause auto-disconnects of the other."
I believe the independent port fuses allows for the independent power regulation for each port as well as protection. I personally went with the Sonnet, but I swap setups regularly and wanted it to speed change-ups for when I need it-because of no additional power needed.; it only peaks at 15w per port though.
Last edited:
Having read stories of the odd PCI card frying some poor soul's Mac Pro motherboard, I'm pretty cautious about what I allow in my case. Maybe it's just marketing, but it "feels like" the Sonnet cards are of a safer, higher-quality product. Regardless, I do know I will not consider risking the health of my treasured 5,1 to a cheap, pizza-priced PCI card shipped direct from China! 
If you ever read about the PCIe to M2 adapter market, you probably will have nightmares… People buy $3~$5 adapters and install $700 NVMe drives…Having read stories of the odd PCI card frying some poor soul's Mac Pro motherboard, I'm pretty cautious about what I allow in my case. Maybe it's just marketing, but it "feels like" the Sonnet cards are of a safer, higher-quality product. Regardless, I do know I will not consider risking the health of my treasured 5,1 to a cheap, pizza-priced PCI card shipped direct from China!
It's not all marketing if it's actually got a warranty; Sonnet has a brand to back up their price-point. That's worth it, to me, when it's a sub-$100 component-IMHO (I got the 2-port).
Anyone currently have a USB 3.1 card that's using an ASM 3142 controller in their Mac Pro? If so, I'd like to know how it's performing for you. It's supposed to have a significant advantage over 1142/2142.
The 3142 has three differences compared to the 2142:
- ASMedia ASM3142 also supports the new feature of USB3.1 Gen2, named Multiple INs to achieve fully utilization rate of 10Gbps data rate, taking advantage of the available bandwidth of USB3.1 Gen2 bus.
- advance chip power management to reduce total power consumption efficiently under idle/standby state.
- Support the Debugport
Only the first one seems significant. If Multiple INs is part of USB 3.1 gen 2, then wouldn't it be part of the ASM1142/ASM2142? I mean, just because the ASM3142 mentions it, doesn't mean the other controllers that don't mention it don't have it also.
It looks like Multiple INs has been in the spec since the start.
https://semiengineering.com/using-usb-3-1s-multiple-ins-to-reach-10-gbps-data-rates/
It seems that Multiple INs is only beneficial when multiple devices are connected, so it shouldn't be an improvement over older controllers if only one device is connected (as is usually the case in benchmarks). A single device can benefit if it has multiple IN endpoints. Do any high bandwidth devices have multiple IN endpoints?
This MSI page says the ASM3142 with Xboost software can do 1140 MB/s which is much closer to the 1212 MB/s max of USB 3.1 gen 2 than I've ever seen. I would like to see benchmarks of a device that can do over 1000 MB/s. I've read that that Xboost will just slow your computer down though.
Not that any additional performance matters since it's still limited to only two PCIe lanes, which in cMP is limited to 750-766MB/s.
There are two ways around this:
1. Find a card with an x4 switch on it, such as the Sonnet Allegro Pro
2. Connect an ASM3142 card to one of the x8 or x16 dual or quad m.2 cards (these have their own PCIe switch)via one of the m.2 slots and a m.2 to PCIe adapter.
Right. You want a card like the Highpoint RocketU 1344A except the two ASM2142 chips are replaced by two ASM3142. The x4 switch on the Sonnet is only PCIe 2.0 for both upstream and downstream links. The Highpoint has a PCIe 3.0 switch to support the downstream ASM2142 at PCIe 3.0 x2. The upstream will operate at PCIe 1.0 or 2.0 or 3.0 speed depending on the slot.
Sonnet is getting ready to release a 3142 card next month. I'd imagine it will have the x4 switch like the Allegro Pro.
Cool. This is not the first PCIe 3.0 card they've made. But it might be the first with a switch? They always list MacPro3,1 compatibility, so I wonder what they'll do about the problem where the switch boots at PCIe 1.0 speed.
Looking to pull my Allegro Pro USB 3.0 card and installing the new Allegro Pro x4 USBC-3.1 Gen 2 in its place.
I am running a 4x4TB RAID5 using the 4 internal bays of my MacPro 5,1. Aja scores it at about 280 write/390 read.
I'm thinking I might pick up a 4 bay USB 3.1 Gen 2 external HDD enclosure such as the OWC Mercury Elite Pro Quad.
If I run the same 4x4TB RAID5 in the new USB3.1 Gen 2 enclosure and card, any guesses what would be a reasonable read/write performance to expect?
FYI-I currently have a 2-bay USB 3.0 enclosure with 2x6TB USB 3 RAID0 connected to the Allegro Pro USB 3.0 card that scores in the 190 write/230 read range. Running Mojave 10.14.3., 3.46 6-core.
I am running a 4x4TB RAID5 using the 4 internal bays of my MacPro 5,1. Aja scores it at about 280 write/390 read.
I'm thinking I might pick up a 4 bay USB 3.1 Gen 2 external HDD enclosure such as the OWC Mercury Elite Pro Quad.
If I run the same 4x4TB RAID5 in the new USB3.1 Gen 2 enclosure and card, any guesses what would be a reasonable read/write performance to expect?
FYI-I currently have a 2-bay USB 3.0 enclosure with 2x6TB USB 3 RAID0 connected to the Allegro Pro USB 3.0 card that scores in the 190 write/230 read range. Running Mojave 10.14.3., 3.46 6-core.
Last edited:
The drives were probably not limited by the 1000 MB/s South Bridge connection or the 300 MB/s SATA ports of the MacPro5,1 so I don't think you should expect much of a change by moving the drives to a 947 MB/s USB enclosure that has 600 MB/s SATA ports. If you switched to SSDs, then you would see an improvement using the USB enclosure.
Are there any mechanical drives that are limited by 300 MB/s SATA ports? Some large 7200 RPM drives can do up to 218MB/s.
Keep in-mind the Sonnet card requires a dedicated 10 Gbps controller to a single port to enable 10 Gbps speeds on the 4,1/5,1. There is only 20 Gbps available, so if only two ports could be used on a four port card - the available bandwidth would be consumed; speeds from there would be compressed dependent upon how many additional ports were used in such case.
joevt is right that SSD's improve the system reactivity but can get expensive if trying to make a large drive like described above.
mavots could expect to potentially double their disk speeds (of the new external array being SATA II limited) if they dedicate a 10 Gbps port to each array, but would be limited to two arrays of full throughput (but USB 3.0 limited for existing array).
Last edited:
Huh? The max theoretical performance of a 10 Gbps connection is 10 Gbps up/10 Gbps down. USB Super Speed and Super Speed+ uses different lines for send and receive - it is full-duplex like PCIe. Only USB Low Speed, Full Speed, and High Speed are half-duplex. SATA has different lines for send and receive but is half-duplex. Half-duplex means they can't send and receive at the same time, but they can still use the full bandwidth each way.
USB Super Speed+ uses 128b/132b encoding so the bandwidth is more like 9.7 Gbps (including USB protocol overhead).
Right. Some SATA III (6 Gbps at 8b/10b = 4.8 Gbps = 600 MB/s) drives are faster than USB Super Speed 5 Gbps (at 8b/10b = 4 Gbps = 500 MB/s) but I think they are all SSDs. Maybe mechanical drives have a cache that is faster than 5 Gbps, but that doesn't matter for sustained reads.
The current Sonnet card has two ASM1142 controllers limited by PCIe 2.0x2 = 5 GT/s * 2 * 8b/10b = 8 Gbps = 1000 MB/s. With PCIe protocol overhead, you might get 750MB/s from a single controller. There is no PCIe 2.0/3.0 conversion. Each controller has two USB ports. If you have two devices, then it's best to connect each device to a different controller.
mavots' current setup maxes out at 390 MB/s which can be achieved from a single USB port. He can't double the speed unless he uses faster hard drives or twice as many hard drives. If he did double the speed, he could almost do it with a single USB port. Doubling the number of hard drives would require an additional array, or an array that supports 8 hard drives.
Last edited:
Got you-but they are talking about use a USB 3.0 enclosure-and I agree on the 1 controller per drive ratio.
On the 4 port card:
"Exploiting its dual controllers design, it’s possible to simultaneously transfer data to two connected drives at up to 660 MB/s each."
http://www.sonnettech.com/product/allegro-usbc-4port-pcie.html
Last edited:
mavots said he was thinking of getting the USB 3.1 gen 2 enclosure.
He did not say that any drives were connected to the old USB 3.0 card that he is wanting to replace with the USB 3.1 gen 2 card.
What description says half-duplex?
Both the old card and the new card are four USB port cards. No USB two port card is mentioned.
I don't know what "x4/2= x2 up/down" means.
And I'm hinting that a two port might save them a few bucks. I own the two port, and am awaiting my 3.1 Gen 2 enclosure for My 6TB in a similar setup-but I don't have it yet due to shipping delays. They list their speeds on their current setup, and if you take that setup and apply it to a new port, you can infer data.
You might want to read on their website as to how those speeds are achieved- I'll let you know how my setup works if/when it gets here though.
Edited: There is lane negotiation taking place.
Last edited: